From: Stan Hilliard
Date: 02 Dec 1999
Time: 22:46:36
Greetings Bob,
I assume that the criterion of <5% errors is an RQL define the consumer's point, and not an AQL. Otherwise there could be a problem. See: www.samplingplans.com/aqlprimer.htm
I have not seen the ANSI/AIIM sampling procedure. Is it available to view on line? The switching rules for tightened and reduced inspection suggest to me that it's AQL based.
You Say: "A set of specially designed PCS will give online access to the images and data for the purpose of sampling." Our software program ASP has the capability of producing a truly random sample of record numbers from a range. See: www.samplingplans.com/programasp.htm
The sample size of 1,000 from a lot of 60,000 seems awfully large. Could the acceptance decision be based on smaller sublots, like dividing the lot into days, or input operator, or microfilm machine? This would make it easier to segregate records in the event of rejection for excessive errors.
You Said: "I would be interested in using suitable software programs if I could be confident that I shall be impementing the correct sampling regime(s)." I think that the thing that makes a sampling regime correct is that you know and accept the probabilities of the sampling plan's oc curve, as shown at: www.samplingplans.com/modern3.htm#EVALUATE
Our software program for developing attribute acceptance plans (TP105) uses the oc curve to develop the plan. It also calculates the oc curve for any n and Ac. It develops in addition to fixed-n plans, sequential sampling plans that make switching rules unnecessary. Sequential sampling varies the sample size according to current quality, as reflected in the current sample rather than the quality of previous lots. See: www.samplingplans.com/programtp105.htm
Sincerely, Stan Hilliard